OV - exec - Loop 95

	Loop Id: 95	Module: exec	Source: timestep.c:86-96	Coverage: 0.04%

0x4107c0 INC	%RAX

0x4107c3 ADD	$0x40,%ESI

0x4107c6 CMP	%RCX,%RAX

0x4107c9 JE	41075b

0x4107cb MOV	(%RDX,%RAX,4),%R11D

0x4107cf TEST	%R11D,%R11D

0x4107d2 JLE	4107c0

0x4107d4 MOV	%ESI,%R9D

0x4107d7 MOV	0x20(%RBX),%R8

0x4107db MOV	0x28(%RBX),%R10

0x4107df MOV	0x10(%R8),%R15

0x4107e3 MOV	0x18(%R8),%RDI

0x4107e7 MOV	0x20(%R8),%R8

0x4107eb CMP	$0x1,%R11D

0x4107ef JE	4108bc

0x4107f5 LEA	(%R9,%R9,2),%R12

0x4107f9 LEA	0x28(,%R12,8),%R12

0x410801 MOV	%R11D,%R13D

0x410804 AND	$0x7ffffffe,%R13D

0x41080b NOPL	(%RAX,%RAX,1)

  (96) 0x410810 MOVSXD	(%R15,%R9,4),%R14

  (96) 0x410814 SAL	$0x4,%R14

  (96) 0x410818 VMOVSD	-0x28(%R8,%R12,1),%XMM1

  (96) 0x41081f VDIVSD	0x8(%R10,%R14,1),%XMM0,%XMM2

  (96) 0x410826 VFMADD213SD	-0x28(%RDI,%R12,1),%XMM2,%XMM1

  (96) 0x41082d VMOVSD	%XMM1,-0x28(%RDI,%R12,1)

  (96) 0x410834 VMOVSD	-0x20(%R8,%R12,1),%XMM1

  (96) 0x41083b VFMADD213SD	-0x20(%RDI,%R12,1),%XMM2,%XMM1

  (96) 0x410842 VMOVSD	%XMM1,-0x20(%RDI,%R12,1)

  (96) 0x410849 VMOVSD	-0x18(%R8,%R12,1),%XMM1

  (96) 0x410850 VFMADD213SD	-0x18(%RDI,%R12,1),%XMM2,%XMM1

  (96) 0x410857 VMOVSD	%XMM1,-0x18(%RDI,%R12,1)

  (96) 0x41085e MOVSXD	0x4(%R15,%R9,4),%R14

  (96) 0x410863 SAL	$0x4,%R14

  (96) 0x410867 VMOVSD	-0x10(%R8,%R12,1),%XMM1

  (96) 0x41086e VDIVSD	0x8(%R10,%R14,1),%XMM0,%XMM2

  (96) 0x410875 VFMADD213SD	-0x10(%RDI,%R12,1),%XMM2,%XMM1

  (96) 0x41087c VMOVSD	%XMM1,-0x10(%RDI,%R12,1)

  (96) 0x410883 VMOVSD	-0x8(%R8,%R12,1),%XMM1

  (96) 0x41088a VFMADD213SD	-0x8(%RDI,%R12,1),%XMM2,%XMM1

  (96) 0x410891 VMOVSD	%XMM1,-0x8(%RDI,%R12,1)

  (96) 0x410898 VMOVSD	(%R8,%R12,1),%XMM1

  (96) 0x41089e VFMADD213SD	(%RDI,%R12,1),%XMM2,%XMM1

  (96) 0x4108a4 VMOVSD	%XMM1,(%RDI,%R12,1)

  (96) 0x4108aa ADD	$0x2,%R9

  (96) 0x4108ae ADD	$0x30,%R12

  (96) 0x4108b2 ADD	$-0x2,%R13D

  (96) 0x4108b6 JNE	410810

0x4108bc TEST	$0x1,%R11B

0x4108c0 JE	4107c0

0x4108c6 MOVSXD	(%R15,%R9,4),%R11

0x4108ca SAL	$0x4,%R11

0x4108ce SAL	$0x3,%R9

0x4108d2 LEA	(%R9,%R9,2),%R9

0x4108d6 VMOVSD	(%R8,%R9,1),%XMM1

0x4108dc VDIVSD	0x8(%R10,%R11,1),%XMM0,%XMM2

0x4108e3 VFMADD213SD	(%RDI,%R9,1),%XMM2,%XMM1

0x4108e9 VMOVSD	%XMM1,(%RDI,%R9,1)

0x4108ef VMOVSD	0x8(%R8,%R9,1),%XMM1

0x4108f6 VFMADD213SD	0x8(%RDI,%R9,1),%XMM2,%XMM1

0x4108fd VMOVSD	%XMM1,0x8(%RDI,%R9,1)

0x410904 VMOVSD	0x10(%R8,%R9,1),%XMM1

0x41090b VFMADD213SD	0x10(%RDI,%R9,1),%XMM2,%XMM1

0x410912 VMOVSD	%XMM1,0x10(%RDI,%R9,1)

0x410919 JMP	4107c0

/home/eoseret/qaas_runs_GNR/173-814-1858/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c: 86 - 96

--------------------------------------------------------------------------------

86:    for (int iBox=0; iBox<nBoxes; iBox++)

87:    {

88:       for (int iOff=MAXATOMS*iBox,ii=0; ii<s->boxes->nAtoms[iBox]; ii++,iOff++)

89:       {

90:          int iSpecies = s->atoms->iSpecies[iOff];

91:          real_t invMass = 1.0/s->species[iSpecies].mass;

92:          s->atoms->r[iOff][0] += dt*s->atoms->p[iOff][0]*invMass;

93:          s->atoms->r[iOff][1] += dt*s->atoms->p[iOff][1]*invMass;

94:          s->atoms->r[iOff][2] += dt*s->atoms->p[iOff][2]*invMass;

95:       }

96:    }

Coverage (%)	Name	Source Location	Module
►100.00+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so

Path /

Metric	Value
CQA speedup if no scalar integer	1.71
CQA speedup if FP arith vectorized	2.31
CQA speedup if fully vectorized	3.42
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.46
Bottlenecks	micro-operation queue,
Function	advancePosition.extracted
Source	timestep.c:86-96
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	6.83
CQA cycles if no scalar integer	4.00
CQA cycles if FP arith vectorized	2.95
CQA cycles if fully vectorized	2.00
Front-end cycles	6.83
P0 cycles	3.50
P1 cycles	3.00
P2 cycles	4.67
P3 cycles	4.67
P4 cycles	1.50
P5 cycles	3.00
P6 cycles	3.50
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	3.00
P11 cycles	4.67
DIV/SQRT cycles	4.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	37.00
Nb uops	37.00
Nb loads	14.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	1.02
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	18.73
Bytes prefetched	0.00
Bytes loaded	104.00
Bytes stored	24.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	10.55
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	6.25
Vector-efficiency ratio fma	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	7.50

Metric	Value
CQA speedup if no scalar integer	1.71
CQA speedup if FP arith vectorized	2.31
CQA speedup if fully vectorized	3.42
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.46
Bottlenecks	micro-operation queue,
Function	advancePosition.extracted
Source	timestep.c:86-96
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	6.83
CQA cycles if no scalar integer	4.00
CQA cycles if FP arith vectorized	2.95
CQA cycles if fully vectorized	2.00
Front-end cycles	6.83
P0 cycles	3.50
P1 cycles	3.00
P2 cycles	4.67
P3 cycles	4.67
P4 cycles	1.50
P5 cycles	3.00
P6 cycles	3.50
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	3.00
P11 cycles	4.67
DIV/SQRT cycles	4.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	37.00
Nb uops	37.00
Nb loads	14.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	1.02
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	18.73
Bytes prefetched	0.00
Bytes loaded	104.00
Bytes stored	24.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	10.55
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	6.25
Vector-efficiency ratio fma	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	7.50

Path /

Average path: Display a virtual path defined by average values of all real paths

Function	advancePosition.extracted
Source file and lines	timestep.c:86-96
Module	exec

The loop is defined in /home/eoseret/qaas_runs_GNR/173-814-1858/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:86-96.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=5
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

Try to simplify control and/or increase the maximum number of paths per function/loop through the 'max-paths-nb' option.

This loop has 5 execution paths.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 6.83 to 4.00 cycles (1.71x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

Vectorization

Your loop is not vectorized. Only 10% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 6.83 to 2.00 cycles (3.42x speedup).

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

FMA

Detected 3 FMA (fused multiply-add) operations.

Type of elements and instruction set

4 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 7 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
3: multiply (all inside FMA instructions)
1: divide

The binary loop is loading 104 bytes (13 double precision FP elements). The binary loop is storing 24 bytes (3 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.05 FP operations per loaded or stored byte.

General properties

nb instructions	37
nb uops	37
loop length	178
used x86 registers	13
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	6.83 cycles
front end	6.83 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67
cycles	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions

4.00

Cycles summary

Front-end	6.83
Dispatch	4.67
DIV/SQRT	4.00
Overall L1	6.83

Vectorization ratios

INT

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	7%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	NA (no fma vectorizable/vectorized instructions)
other	7%

all	12%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	12%
div/sqrt	12%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	10%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	12%
div/sqrt	12%
other	7%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 6.83 cycles. At this rate:

7% of peak load performance is reached (15.22 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (3.51 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 6.83 to 4.67 cycles (1.46x speedup).

ASM code

In the binary file, the address of the loop is: 4107cb

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x40,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
CMP %RCX,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JE 41075b <advancePosition.extracted+0x6b>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV (%RDX,%RAX,4),%R11D	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
TEST %R11D,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	scal (6.3%)
JLE 4107c0 <advancePosition.extracted+0xd0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV %ESI,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV 0x20(%RBX),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x28(%RBX),%R10	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x10(%R8),%R15	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x18(%R8),%RDI	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x20(%R8),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
CMP $0x1,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
JE 4108bc <advancePosition.extracted+0x1cc>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (%R9,%R9,2),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA 0x28(,%R12,8),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %R11D,%R13D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%R13D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NOPL (%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
TEST $0x1,%R11B	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	N/A
JE 4107c0 <advancePosition.extracted+0xd0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOVSXD (%R15,%R9,4),%R11	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
SAL $0x4,%R11	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
SAL $0x3,%R9	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
LEA (%R9,%R9,2),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VMOVSD (%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VDIVSD 0x8(%R10,%R11,1),%XMM0,%XMM2	1	1	0	0.33	0.33	0	0	0	0	0	0	0	0.33	13-15	4	scal (12.5%)
VFMADD213SD (%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x8(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x8(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x8(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x10(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x10(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x10(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
JMP 4107c0 <advancePosition.extracted+0xd0>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A

Function	advancePosition.extracted
Source file and lines	timestep.c:86-96
Module	exec

The loop is defined in /home/eoseret/qaas_runs_GNR/173-814-1858/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:86-96.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=5
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 6.83 to 4.00 cycles (1.71x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

Vectorization

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

FMA

Detected 3 FMA (fused multiply-add) operations.

Type of elements and instruction set

4 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 7 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
3: multiply (all inside FMA instructions)
1: divide

The binary loop is loading 104 bytes (13 double precision FP elements). The binary loop is storing 24 bytes (3 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.05 FP operations per loaded or stored byte.

General properties

nb instructions	37
nb uops	37
loop length	178
used x86 registers	13
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	6.83 cycles
front end	6.83 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67
cycles	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions

4.00

Cycles summary

Front-end	6.83
Dispatch	4.67
DIV/SQRT	4.00
Overall L1	6.83

Vectorization ratios

INT

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	7%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	NA (no fma vectorizable/vectorized instructions)
other	7%

all	12%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	12%
div/sqrt	12%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	10%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	12%
div/sqrt	12%
other	7%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 6.83 cycles. At this rate:

7% of peak load performance is reached (15.22 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (3.51 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 4107cb

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x40,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
CMP %RCX,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JE 41075b <advancePosition.extracted+0x6b>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV (%RDX,%RAX,4),%R11D	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
TEST %R11D,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	scal (6.3%)
JLE 4107c0 <advancePosition.extracted+0xd0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV %ESI,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV 0x20(%RBX),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x28(%RBX),%R10	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x10(%R8),%R15	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x18(%R8),%RDI	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x20(%R8),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
CMP $0x1,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
JE 4108bc <advancePosition.extracted+0x1cc>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (%R9,%R9,2),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA 0x28(,%R12,8),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %R11D,%R13D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%R13D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NOPL (%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
TEST $0x1,%R11B	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	N/A
JE 4107c0 <advancePosition.extracted+0xd0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOVSXD (%R15,%R9,4),%R11	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
SAL $0x4,%R11	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
SAL $0x3,%R9	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
LEA (%R9,%R9,2),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VMOVSD (%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VDIVSD 0x8(%R10,%R11,1),%XMM0,%XMM2	1	1	0	0.33	0.33	0	0	0	0	0	0	0	0.33	13-15	4	scal (12.5%)
VFMADD213SD (%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x8(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x8(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x8(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x10(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x10(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x10(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
JMP 4107c0 <advancePosition.extracted+0xd0>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Type of elements and instruction set

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Type of elements and instruction set

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code